Mail Archives: geda-user/2012/11/16/03:59:49
On Fri, Nov 16, 2012 at 03:39:46AM +0100, Kai-Martin Knaak wrote:
> "Peter C.J. Clifton" <pcjc2 AT cam DOT ac DOT uk> schrieb am 14. November 2012:
>
> > This probably won't play well with the renumber tools out there,
>
> Specifically, it does not play well with the built-in renumber tool of
> gschem. Attributes -> autonumber-text will always make all refdeses
> different even if they were deliberately set to the same value. This is
> one of my "favourite" gschem warts.
Agreed. But do you have a plan to solve this problem?
>
>
> > Finally.. if you do use multiple symbols anywher, there is (or was?) a
> > problem you have to work around, in that when picking up the symbols
> > to retrieve attributes from (e.g. for forwarding to PCB layout), it
> > was not clear which of the two symbols would be scanned for attributes.
> >
> > To get around this, add the "footprint=..." and any other attributes
> > you need to be "machine" readable to both halves of the symbol.
>
> This kind of redundancy creates maintenance issues. E.g. consider a
> 7414, which is a package with six inverters. In the schematic these
> would be six times a NOT symbol plus a symbol for the power pins. All of
> them have to be edited if the footprint is to be changed from DIP14 to
> SO14.
That's a problem.
> I like to make the footprint value visible in the schematics.
On that point we disagree.
> The requirement for footprint attributes on each and every symbol adds
> to the clutter in the schematic.
>
>
> > (If this issue hase has been resolved already, someone please let me
> > know).
>
> It has been alleviated, but not quite resolved. Current gnetlist warns
> if it finds different footprints to an attribute. However:
>
> * which value gnetlist picks, depends on the order the symbols were
> entered in the schematic. I wrote a patch that makes sure, gnetlist
> output never ever depends on the order of symbols. This patch was not
> accepted by the devs. I still think, this dependence on order is a bug.
I agree, especially when the attribute is only found once. Conflicting
attributes for the same refdes are a different issue and should always
elicit a warning, at least if they are used in netlist/simulation (not
the "comment" attribute for example).
In the case of a 7414, the logical way is to assign the footprint to the
power symbol and only to it.
>
> * import-schematics in pcb does not show the warnings
>
> * a missing footprint attribute is treated like a footprint with value
> "#f". This effectively enforces to set the footprint attribute on
> every symbol. Else, gnetlist output will be swamped with warnings.
That one should not fixable, although it may be hard. One thing I dislike
is that gnetlist returns "unknown" in attribute getters in scheme when
it should be #f IMHO.
I once dived into gnetlist code and got lost. I would need to study it more
but won't have time in the foreseeable future.
Gabriel
- Raw text -